Analysis of Discrete Variables in Animal Breeding Contexts

Abstract

Many traits of bioeconomic importance in animal production (survival, calving difficulty and disease resistance) present discontinuous distributions of phenotypes. Animal breeders generally have resorted to conventional techniques based on the multivariate normal distribution to estimate functions and test hypotheses on discrete data. The linear logistic model is an alternative to conventional techniques for fixed models describing dichotomous and polychotomous random variables. Data from a line-crossing experiment with rats illustrate computational procedures for the logistic model and this is compared against ordinary least-squares techniques.